K2+PFS, a Beautiful Exoplanet Characterizing Partnership

This is post is mostly by Liang, with a little bit by Johanna at the end.

Johanna said to include a picture of me "doing science". For most astronomers,
most of the time, this is what "doing science" looks like!

My name is Liang Yu and I'm a second-year graduate student at MIT. I started working on the K2 mission about five months ago. Since it began in 2013, the mission has yielded about 300 transiting exoplanet candidates. I'd like to share with you some of these exciting results and a behind-the-scenes sneak peek of how we extract signals from planet candidates from large quantities of K2 data. Thank you Johanna for starting this wonderful blog!

K2 is the "successor" of NASA's highly successful Kepler mission, which has discovered thousands of transiting exoplanet candidates. By staring at stars in one patch of sky and watching for them to change brightness, Kepler could detect when a planet passed in front of a star and blocked out a tiny fraction of its light. (If you want to try detecting Kepler planets yourself, try out this simulation or this Citizen Science project!) By May 2013, two of the spacecraft's four reaction wheels had failed, leaving Kepler unable to point precisely at its original field (in the Cygnus constellation). Luckily, the spacecraft is still able to balance itself against solar radiation pressure and continue taking data, and the mission has since been renamed "K2". K2 is different in that it does not point at one patch of sky continuously, but observes different fields of stars for 80 days each. This means it has a shorter baseline for catching stellar brightness changes due to planets (so, it cannot detect longer period planets), but also that we now get to search for transiting planets all across the sky. Another different and cool thing about K2 is that it is almost totally community-driven -- scientists propose for what targets they want observed in each field, and the data are public immediately, so anyone with a computer can analyze them. 

K2 delivers images of each field from which we can produce a time series of each star's brightness level (called a light curve). But astronomers quickly discovered that the reduced pointing precision of K2 introduced significant levels of systematic noise into the light curves. Without any special processing, most transit signals would be hidden by the noise.

This is what a raw image of a star looks like, straight from K2. The star can sometimes look elongated due to the spacecraft's motion.

There is a fix for this problem, however. Systematic fluctuations in the light curves are strongly correlated with the camera's motion. The spacecraft slowly drifts and is straightened by thruster fires every 6 hours, leading to sudden jumps in the light curves. After several rounds of cleaning that corrects for systematic noise as a function of the spacecraft's motion, and removing the low-frequency variability caused by stellar activity, we get much higher quality data that can reveal transits shallower (smaller) than the level of systematic noise. That translates to smaller planets! 

Top: The uncorrected light curve from K2. Bottom: The same light curve after removing noise correlated to
the spacecraft motion and stellar activity. Now we can clearly see the little dips that may correspond
to a planet passing in front of the star at regular intervals.

After running this procedure on tens of thousands of target stars in each K2 field (which usually takes at least a few hours), we then run a procedure called Box-Least-Squares on all the cleaned light curves to find the ones with periodic dips that look like exoplanets eclipsing their host stars. Folding each light curve so that the dips overlap, we get these (sometimes) beautiful plots of exoplanet transits. The shapes and sizes of these dips can reveal a wealth of information about the planets causing the transits.

A cleaned and folded light curve, fitted with a theoretical model of a transit (red curve).

First of all, the depth of the transit tells us the size of the planet candidate relative to its host star. The deeper the transit, the larger the planet compared to the star. Combining this information with the absolute size of the star, which we can estimate from images taken at different wavelengths, we can derive the absolute radius of the planet. Previous work has uncovered an empirical relation between planetary radius and mass, which works well at least for small planets. Thus we can plug our estimate of the planets' radius into this relation to get a first guess for their masses. However, to get a more reliable estimate, we must turn to ground-based follow-up observations.

That's where the Planet Finder Spectrograph (PFS) at Magellan II comes in! I've written about PFS before, but not much about the science it actually accomplishes. PFS detects planets via the "Doppler wobble" or radial velocity technique -- we observe stars and spread out their light through a grating, like sunlight through a prism, into the different components or wavelengths of the light. As light escapes a star's interior through it's atmosphere, the atmosphere absorbs a bit of the light in specific patterns, depending on what the atmosphere is made of. This is observable in lines "missing" from the star's spectrum. In addition to the composition of the atmosphere, the motion of the star (moving towards or away from us) also controls the positioning of the absorption lines.

In a simplistic picture, we can think of stars as stable, non-moving objects. In reality, they have their own orbits and motions around the Galaxy, are sometimes interacting with each other, and our Sun is moving relative to them. But if we just observe one star and compare it to itself at some reference point, the motion should be pretty small. And yet, we see evidence of stars moving more than expected in the shifting positions of the absorption lines in their spectra. It appears that they are moving towards and away from us in a small but measurable and very periodic way. What could be causing this motion?

PLANETS!

 

This is a video from ESO (L. Calçada), showing how a star's motion towards and away from us is 1) reflected in the lines of its spectrum and 2) caused by a planet orbiting the star and causing the motion. We are a little competitive with ESO (their precision radial velocity spectrograph is called HARPS), but they make a mean movie so I couldn't help myself.

While in most cases stars are much(x3) more massive than planets, they still obey the laws of gravity, so that the star effectively feels a tiny gravitational pull from the planet and the planet effectively feels a much larger gravitational pull from the star. From afar, it seems like the planet is just orbiting the star, but in reality (as is shown in an exaggerated way in the video), the star and planet are both orbiting a common center of mass. With PFS, we distinguish tiny shifts in the star's absorption lines, caused by tiny tugs from planets, by comparing how these lines move relative to an actually-stable reference spectrum, typically of iodine. We superimpose the iodine spectrum on top of the star's lines, and thus have "zero point" around which we can measure how the star's lines change. 

Carnegie DTM's Paul Butler helped pioneer this iodine-cell radial velocity (RV) technique to detect some of the first exoplanets ever. In fact, the 20th anniversary of the first exoplanet detected this way around a solar-like star, 51 Peg b, was this year. For a long time, RV was the most productive way to detect exoplanets; it really wasn't until Kepler was launched that the RV planet detection rate was surpassed. 

The beautiful thing is, both RV and transit techniques are really needed to learn about planets, since the first only measures a mass and the second only measures a radius. If we want to understand what planets are made of -- and we very much do! -- we need both techniques. This has led to a growing partnership between Kepler and K2 planet-finders and ground-based RV observers, one of which is now K2+PFS. We also partner with the HAT-South transiting planet survey, and have already helped confirm and better characterize ~10 transiting exoplanets (that is in addition to the ~10 RV-only planets PFS has detected or help confirm). One of my favorite transiting planet follow-ups so far is WASP-97, an unusual system because it hosts a small, probably-rocky planet inside the orbit of a giant, hot Jupiter planet; the smaller planet goes around the star every ~19 hours.  PFS measured or helped confirm the masses and orbits of all of the planets in this system. 

Mind blown, right?